Object detection in image based on stochastic optimization

Abstract

An electronic device includes circuitry that determines probability map information for a first image, based on application of a neural network model on the first image. The neural network model is trained to detect one or more objects based on a plurality of images associated with the one or more objects. The probability map information indicates a probability value for each pixel in the first image. A region corresponding to the one or more objects is detected in the first image based on the probability map information. A first set of sub-images is determined from the detected region, based on application of a stochastic optimization function on the probability map information. The one or more objects are detected from a second set of sub-images of the first set of sub-images, based on application of the neural network model on the second set of sub-images.

What is claimed is: 
     
       1. An electronic device, comprising:
 circuitry configured to:
 determine probability map information for a first image of a first size, based on application of a neural network model on the first image, wherein
 the neural network model is trained to detect at least one object based on a plurality of images associated with the at least one object, 
 the determined probability map information indicates a probability value for each pixel of a plurality of pixels of the first image, and 
 the plurality of pixels is associated with the at least one object in the first image; 
 
 detect a region that corresponds to the at least one object in the first image based on the determined probability map information of the first image, wherein
 the detected region includes a set of pixels of the plurality of pixels, and 
 the probability value for each pixel of the set of pixels exceeds a threshold value; 
 
 determine a first set of sub-images from the detected region, based on application of a stochastic optimization function on the determined probability map information, wherein a second size of each of the first set of sub-images is less than the first size of the first image; 
 determine at least one of a number of sub-images of the first set of sub-images, the second size of each of the first set of sub-images, or a position of each of the first set of sub-images in the first image, based on the application of the stochastic optimization function on the determined probability map information; 
 crop the first set of sub-images from the first image based on the determined at least one of the number of sub-images of the first set of sub-images, the second size of each of the first set of sub-images, or the position of each of the first set of sub-images in the first image; 
 select a second set of sub-images from the cropped first set of sub-images; and 
 detect the at least one object from the second set of sub-images of the first set of sub-images, based on application of the neural network model on the second set of sub-images. 
 
 
     
     
       2. The electronic device according to  claim 1 , further comprising an image capturing device configured to capture the first image, wherein the circuitry is further configured to:
 detect a change in an imaging parameter associated with the image capturing device; and 
 determine the probability map information based on the detected change in the imaging parameter. 
 
     
     
       3. The electronic device according to  claim 2 , wherein the imaging parameter associated with the image capturing device comprises at least one of a position parameter associated with the image capturing device, an orientation parameter associated with the image capturing device, a zooming parameter associated with the image capturing device, a type of an image sensor associated with the image capturing device, a pixel size associated with the image sensor of the image capturing device, a lens type associated with the image capturing device, a focal length associated with the image capturing device to capture the first image, or a geo-location of the image capturing device. 
     
     
       4. The electronic device according to  claim 2 , wherein the circuitry is further configured to:
 control the image capturing device to capture a second image at a specific time interval; and 
 determine the probability map information for the captured second image. 
 
     
     
       5. The electronic device according to  claim 1 , wherein the at least one object corresponds to a license plate of at least one vehicle in the first image. 
     
     
       6. The electronic device according to  claim 1 , wherein the neural network model comprises one of an artificial neural network (ANN), a convolutional neural network (CNN), a CNN-recurrent neural network (CNN-RNN), Region-CNN (R-CNN), Fast R-CNN, Faster R-CNN, a Long Short Term Memory (LSTM) network based RNN, a combination of CNN and ANN, a combination of LSTM and ANN, a gated recurrent unit (GRU)-based RNN, a deep Bayesian neural network, a Generative Adversarial Network (GAN), a deep learning based object detection model, a feature-based object detection model, an image segmentation based object detection model, a blob analysis-based object detection model, a “you look only once” (YOLO) object detection model, or a single-shot multi-box detector (SSD) based object detection model. 
     
     
       7. The electronic device according to  claim 1 , wherein the stochastic optimization function comprises one of a cost function, a direct search function, a simultaneous perturbation function, a Hill climbing function, a random search function, a Tabu search function, a Particle Swarm Optimization (PSO) function, an Ant Colony Optimization function, a simulated annealing function, or a genetic function. 
     
     
       8. The electronic device according to  claim 1 , wherein the circuitry is further configured to:
 detect at least one bounding box in the first image based on the application of the neural network model on the first image, wherein the at least one bounding box includes the at least one object; and 
 determine the probability map information for the first image of the first size based on the detected at least one bounding box in the first image. 
 
     
     
       9. A method, comprising:
 in an electronic device:
 determining probability map information for a first image of a first size, based on application of a neural network model on the first image, wherein
 the neural network model is trained to detect at least one object based on a plurality of images associated with the at least one object, 
 the determined probability map information indicates a probability value for each pixel of a plurality of pixels of the first image, and 
 the plurality of pixels is associated with the at least one object in the first image; 
 
 detecting a region that corresponds to the at least one object in the first image based on the determined probability map information of the first image, wherein
 the detected region includes a set of pixels of the plurality of pixels, and 
 the probability value for each pixel of the set of pixels exceeds a threshold value; 
 
 determining a first set of sub-images from the detected region, based on application of a stochastic optimization function on the determined probability map information, wherein a second size of each of the first set of sub-images is less than the first size of the first image; 
 determining at least one of a number of sub-images of the first set of sub-images, the second size of each of the first set of sub-images, or a position of each of the first set of sub-images in the first image, based on the application of the stochastic optimization function on the determined probability map information; 
 cropping the first set of sub-images from the first image based on the determined at least one of the number of sub-images of the first set of sub-images, the second size of each of the first set of sub-images, or the position of each of the first set of sub-images in the first image; 
 selecting a second set of sub-images from the cropped first set of sub-images; and 
 detecting the at least one object from the second set of sub-images of the first set of sub-images, based on application of the neural network model on the second set of sub-images. 
 
 
     
     
       10. The method according to  claim 9 , further comprising:
 detecting a change in an imaging parameter associated with an image capturing device of the electronic device; and 
 determining the probability map information based on the detected change in the imaging parameter. 
 
     
     
       11. The method according to  claim 10 , wherein the imaging parameter associated with the image capturing device comprises at least one of a position parameter associated with the image capturing device, an orientation parameter associated with the image capturing device, a zooming parameter associated with the image capturing device, a type of an image sensor associated with the image capturing device, a pixel size associated with the image sensor of the image capturing device, a lens type associated with the image capturing device, a focal length associated with the image capturing device to capture the first image, or a geo-location of the image capturing device. 
     
     
       12. The method according to  claim 10 , further comprising:
 controlling the image capturing device to capture a second image at a specific time interval; and 
 determining the probability map information for the captured second image. 
 
     
     
       13. The method according to  claim 9 , wherein the at least one object corresponds to a license plate of at least one vehicle in the first image. 
     
     
       14. The method according to  claim 9 , wherein the neural network model comprises one of an artificial neural network (ANN), a convolutional neural network (CNN), a CNN-recurrent neural network (CNN-RNN), Region-CNN (R-CNN), Fast R-CNN, Faster R-CNN, a Long Short Term Memory (LSTM) network based RNN, a combination of CNN and ANN, a combination of LSTM and ANN, a gated recurrent unit (GRU)-based RNN, a deep Bayesian neural network, a Generative Adversarial Network (GAN), a deep learning based object detection model, a feature-based object detection model, an image segmentation based object detection model, a blob analysis-based object detection model, a “you look only once” (YOLO) object detection model, or a single-shot multi-box detector (SSD) based object detection model. 
     
     
       15. The method according to  claim 9 , wherein the stochastic optimization function comprises one of a cost function, a direct search function, a simultaneous perturbation function, a Hill climbing function, a random search function, a Tabu search function, a Particle Swarm Optimization (PSO) function, an Ant Colony Optimization function, a simulated annealing function, or a genetic function. 
     
     
       16. The method according to  claim 9 , further comprising:
 detecting at least one bounding box in the first image based on the application of the neural network model on the first image, wherein the at least one bounding box includes the at least one object; and 
 determining the probability map information for the first image of the first size based on the detected at least one bounding box in the first image. 
 
     
     
       17. A non-transitory computer-readable medium having stored thereon, computer-executable instructions that when executed by an electronic device, causes the electronic device to execute operations, the operations comprising:
 determining probability map information for an image of a first size, based on application of a neural network model on the image, wherein
 the neural network model is trained to detect at least one object based on a plurality of images associated with the at least one object, 
 the determined probability map information indicates a probability value for each pixel of a plurality of pixels of the image, and 
 the plurality of pixels is associated with the at least one object in the image; 
 
 detecting a region that corresponds to the at least one object in the image based on the determined probability map information of the image, wherein
 the detected region includes a set of pixels of the plurality of pixels, and 
 the probability value for each pixel of the set of pixels exceeds a threshold value; 
 
 determining a first set of sub-images from the detected region, based on application of a stochastic optimization function on the determined probability map information, wherein a second size of each of the first set of sub-images is less than the first size of the image; 
 determining at least one of a number of sub-images of the first set of sub-images, the second size of each of the first set of sub-images, or a position of each of the first set of sub-images in the image, based on the application of the stochastic optimization function on the determined probability map information; 
 cropping the first set of sub-images from the image based on the determined at least one of the number of sub-images of the first set of sub-images, the second size of each of the first set of sub-images, or the position of each of the first set of sub-images in the image; 
 selecting a second set of sub-images from the cropped first set of sub-images; and 
 detecting the at least one object from a second set of sub-images of the first set of sub-images, based on application of the neural network model on the second set of sub-images. 
 
     
     
       18. The non-transitory computer-readable medium according to  claim 17 , wherein the at least one object corresponds to a license plate of at least one vehicle in the image.